Learning olfactory codes using matrix factorization on 2DG uptake patterns from rats
Flavour volume 3, Article number: O3 (2014)
Recent studies based on both the analysis of odor descriptors  and psychophysical studies  have suggested that olfactory perception is mainly dominated by odor pleasantness. In contrast, findings on the sensory side  as well as on the perceptual side  indicate an odor perception space that is spanned by about 30 dimensions. On the way to understand the basic principles in the decoding of olfactory information from the single molecular compounds, we are focusing in this study on the first level of information processing in the sense of smell: The olfactory bulb (OB). We know that the olfactory epithelium is separated into four regions  of olfactory receptor protein expression and we expect a similar organization in the olfactory bulb . With 2-DG-uptake-images, Johnson et al.  measured odorant-specific activity patterns, covering 98% of the bulbar surface with a resolution of one glomerulus per pixel. Using a database of 143 such bulbar activity fingerprints, we propose the so called non-negative matrix factorization  to decompose the data into a set of common features. These features can be best described as the coding alphabet while the uptake images of single molecular compounds resemble natural occurring code words. Interestingly, we can achieve three meaningful alphabets with a different size: 4, 6, and 9 digits. Only the last one is suitable for actually coding the compounds, as a single compound is evoking activity of only about 1/8 of the bulbar surface, with an average code word length of two to three digits. The other two decompositions might reflect the superglomerular organization of the receptor neurons in the bulb, like e.g. the above mentioned expression patterns and neuronal subnetworks that directly result from the targeting mechanisms guiding the neurons from the OE to the OB. We argue that the proposed decompositions are not only reflecting the chemotopic organization of the bulbus but rather giving a first hint on the coding scheme underlying olfactory perception.
Madany Mamlouk A, Chee-Ruiter C, Hofmann UG, Bower JM: Quantifying olfactory perception: mapping olfactory perception space by using multidimensional scaling and self-organizing maps. Neurocomputing. 2003, 52-54: 591-597.
Khan RM, Luk C-H, Flinker A, Aggarwal A, Lapid H, Haddad R, Sobel N: Predicting odor pleasantness from odorant structure: pleasantness as a reflection of the physical world. J Neurosci. 2007, 27: 10015-23.
Haddad R, Khan R, Takahashi Y, Mori K, Harel D, Sobel N: A metric for odorant comparison. Nat Methods. 2008, 5: 425-429.
Madany Mamlouk A, Martinetz T: On the dimensions of the olfactory perception space. Neurocomputing. 2004, 58-60: 1019-1025.
Ressler KJ, Sullivan SL, Buck LB: A zonal organization of odorant receptor gene expression in the olfactory epithelium. Cell. 1993, 73: 597-609.
Mori K, Takahashi YK, Igarashi KM, Yamaguchi M: Maps of odorant molecular features in the Mammalian olfactory bulb. Physiol Rev. 2006, 86: 409-33.
Johnson Ba, Woo CC, Leon M: Spatial coding of odorant features in the glomerular layer of the rat olfactory bulb. J Comp Neurol. 1998, 393: 457-71.
Lee DD, Seung HS: Learning the parts of objects by non-negative matrix factorization. Nature. 1999, 401: 788-91.
About this article
Cite this article
Mamlouk, A.M. Learning olfactory codes using matrix factorization on 2DG uptake patterns from rats. Flavour 3 (Suppl 1), O3 (2014). https://doi.org/10.1186/2044-7248-3-S1-O3